Estimating And Application Of CVaR Based On The AR-GARCH-EVT Model

In this paper,for the lack of the additivity,convexity and etc in traditional risk measurement methods, we introduces a new financial risk measurement tool:CVaR. It makes up for the VaR to do nothing about the part of over the threshold point (VaR) Relative to the research of dynamic VaR,there are two aspects about the analysis of CVaR:on the one hand, it is the risk measurement model, which usually assumes that the profit and loss distribution is normal distribution, but it ignores the actual financial data showing a spike in thick tail and extreme; on the other hand,from the volatility of actual financial data, we can build appropriate volatility model.We solve the clustering of financial time series volatility by GARCH models, and calculate the dynamic condition means and variances of historical data. For the advantage of the extreme value theory to estimate fat-tailed distribution, this paper deal with the tail end of the financial assets yield sequence using the POT model. Then we introduce the concept of CVaR, and use the combination of GARCH models, POT and CVaR model in order to finally put forward the general AR-GARCH-EVT CVaR model in the paper.For testing the practicality and effectiveness of this model, this article make the actual test and verify by fitting the Shanghai index3132historical data. Through calculation and testing failure frequency, and comparing the results with the tradi-tional VaR model, we find that the proposed method of our paper gives a better1-day CVaR-estimation than traditional VaR estimated.